diff options
Diffstat (limited to 'libffi/src/loongarch64/sysv.S')
-rw-r--r-- | libffi/src/loongarch64/sysv.S | 327 |
1 files changed, 327 insertions, 0 deletions
diff --git a/libffi/src/loongarch64/sysv.S b/libffi/src/loongarch64/sysv.S new file mode 100644 index 0000000..aa7bde2 --- /dev/null +++ b/libffi/src/loongarch64/sysv.S @@ -0,0 +1,327 @@ +/* ----------------------------------------------------------------------- + sysv.S - Copyright (c) 2022 Xu Chenghua <xuchenghua@loongson.cn> + 2022 Cheng Lulu <chenglulu@loongson.cn> + + LoongArch Foreign Function Interface + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + ``Software''), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice shall be included + in all copies or substantial portions of the Software. + + THE SOFTWARE IS PROVIDED ``AS IS'', WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT + HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + DEALINGS IN THE SOFTWARE. + ----------------------------------------------------------------------- */ + +#define LIBFFI_ASM +#include <fficonfig.h> +#include <ffi.h> + +/* Define aliases so that we can handle all ABIs uniformly. */ + +#if __SIZEOF_POINTER__ == 8 +# define PTRS 8 +# define LARG ld.d +# define SARG st.d +#else +# define PTRS 4 +# define LARG ld.w +# define SARG st.w +#endif + +#if defined(__loongarch_single_float) +# define FLTS 4 +# define FLD fld.w +# define FST fst.w +#elif defined(__loongarch_double_float) +# define FLTS 8 +# define FLARG fld.d +# define FSARG fst.d +#elif defined(__loongarch_soft_float) +# define FLTS 0 +#else +#error unsupported LoongArch floating-point ABI +#endif + + .text + .globl ffi_call_asm + .type ffi_call_asm, @function + .hidden ffi_call_asm +/* struct call_context + { + ABI_FLOAT fa[8]; + size_t a[10]; + } + + - 8 floating point parameter/result registers (fa[0] - fa[7]) + - 8 integer parameter/result registers (a[0] - a[7]) + - 2 registers used by the assembly code to in-place construct its own stack + frame. + - frame pointer (a[8]) + - return address (a[9]) + + void ffi_call_asm (size_t *stackargs, struct call_context *regargs, + void (*fn)(void), void *closure); */ + +#define FRAME_LEN (8 * FLTS + 10 * PTRS) + +ffi_call_asm: + .cfi_startproc + + /* We are NOT going to set up an ordinary stack frame. In order to pass + the stacked args to the called function, we adjust our stack pointer + to a0, which is in the _caller's_ alloca area. We establish our own + stack frame at the end of the call_context. + + Anything below the arguments will be freed at this point, although + we preserve the call_context so that it can be read back in the + caller. */ + + .cfi_def_cfa 5, FRAME_LEN # Interim CFA based on a1. + SARG $fp, $a1, FRAME_LEN - 2*PTRS + .cfi_offset 22, -2*PTRS + SARG $ra, $a1, FRAME_LEN - 1*PTRS + .cfi_offset 1, -1*PTRS + + addi.d $fp, $a1, FRAME_LEN + move $sp, $a0 + .cfi_def_cfa 22, 0 # Our frame is fully set up. + + # Load arguments. + move $t1, $a2 + move $t2, $a3 + +#if FLTS + FLARG $fa0, $fp, -FRAME_LEN+0*FLTS + FLARG $fa1, $fp, -FRAME_LEN+1*FLTS + FLARG $fa2, $fp, -FRAME_LEN+2*FLTS + FLARG $fa3, $fp, -FRAME_LEN+3*FLTS + FLARG $fa4, $fp, -FRAME_LEN+4*FLTS + FLARG $fa5, $fp, -FRAME_LEN+5*FLTS + FLARG $fa6, $fp, -FRAME_LEN+6*FLTS + FLARG $fa7, $fp, -FRAME_LEN+7*FLTS +#endif + + LARG $a0, $fp, -FRAME_LEN+8*FLTS+0*PTRS + LARG $a1, $fp, -FRAME_LEN+8*FLTS+1*PTRS + LARG $a2, $fp, -FRAME_LEN+8*FLTS+2*PTRS + LARG $a3, $fp, -FRAME_LEN+8*FLTS+3*PTRS + LARG $a4, $fp, -FRAME_LEN+8*FLTS+4*PTRS + LARG $a5, $fp, -FRAME_LEN+8*FLTS+5*PTRS + LARG $a6, $fp, -FRAME_LEN+8*FLTS+6*PTRS + LARG $a7, $fp, -FRAME_LEN+8*FLTS+7*PTRS + + /* Call */ + jirl $ra, $t1, 0 + +#if FLTS + /* Save return values - only a0/a1 (fa0/fa1) are used. */ + FSARG $fa0, $fp, -FRAME_LEN+0*FLTS + FSARG $fa1, $fp, -FRAME_LEN+1*FLTS +#endif + + SARG $a0, $fp, -FRAME_LEN+8*FLTS+0*PTRS + SARG $a1, $fp, -FRAME_LEN+8*FLTS+1*PTRS + + /* Restore and return. */ + addi.d $sp, $fp, -FRAME_LEN + .cfi_def_cfa 3, FRAME_LEN + LARG $ra, $fp, -1*PTRS + .cfi_restore 1 + LARG $fp, $fp, -2*PTRS + .cfi_restore 22 + jr $ra + .cfi_endproc + .size ffi_call_asm, .-ffi_call_asm + + +/* ffi_closure_asm. Expects address of the passed-in ffi_closure in t0. + void ffi_closure_inner (ffi_cif *cif, + void (*fun)(ffi_cif *, void *, void **, void *), + void *user_data, + size_t *stackargs, struct call_context *regargs) */ + + .globl ffi_closure_asm + .hidden ffi_closure_asm + .type ffi_closure_asm, @function + +ffi_closure_asm: + .cfi_startproc + addi.d $sp, $sp, -FRAME_LEN + .cfi_def_cfa_offset FRAME_LEN + + /* Make a frame. */ + SARG $fp, $sp, FRAME_LEN - 2*PTRS + .cfi_offset 22, -2*PTRS + SARG $ra, $sp, FRAME_LEN - 1*PTRS + .cfi_offset 1, -1*PTRS + addi.d $fp, $sp, FRAME_LEN + + /* Save arguments. */ +#if FLTS + FSARG $fa0, $sp, 0*FLTS + FSARG $fa1, $sp, 1*FLTS + FSARG $fa2, $sp, 2*FLTS + FSARG $fa3, $sp, 3*FLTS + FSARG $fa4, $sp, 4*FLTS + FSARG $fa5, $sp, 5*FLTS + FSARG $fa6, $sp, 6*FLTS + FSARG $fa7, $sp, 7*FLTS +#endif + + SARG $a0, $sp, 8*FLTS+0*PTRS + SARG $a1, $sp, 8*FLTS+1*PTRS + SARG $a2, $sp, 8*FLTS+2*PTRS + SARG $a3, $sp, 8*FLTS+3*PTRS + SARG $a4, $sp, 8*FLTS+4*PTRS + SARG $a5, $sp, 8*FLTS+5*PTRS + SARG $a6, $sp, 8*FLTS+6*PTRS + SARG $a7, $sp, 8*FLTS+7*PTRS + + /* Enter C */ + LARG $a0, $t0, FFI_TRAMPOLINE_SIZE+0*PTRS + LARG $a1, $t0, FFI_TRAMPOLINE_SIZE+1*PTRS + LARG $a2, $t0, FFI_TRAMPOLINE_SIZE+2*PTRS + addi.d $a3, $sp, FRAME_LEN + move $a4, $sp + + bl ffi_closure_inner + + /* Return values. */ +#if FLTS + FLARG $fa0, $sp, 0*FLTS + FLARG $fa1, $sp, 1*FLTS +#endif + + LARG $a0, $sp, 8*FLTS+0*PTRS + LARG $a1, $sp, 8*FLTS+1*PTRS + + /* Restore and return. */ + LARG $ra, $sp, FRAME_LEN-1*PTRS + .cfi_restore 1 + LARG $fp, $sp, FRAME_LEN-2*PTRS + .cfi_restore 22 + addi.d $sp, $sp, FRAME_LEN + .cfi_def_cfa_offset 0 + jr $ra + .cfi_endproc + .size ffi_closure_asm, .-ffi_closure_asm + +/* Static trampoline code table, in which each element is a trampoline. + + The trampoline clobbers t0 and t1, but we don't save them on the stack + because our psABI explicitly says they are scratch registers, at least for + ELF. Our dynamic trampoline is already clobbering them anyway. + + The trampoline has two parameters - target code to jump to and data for + the target code. The trampoline extracts the parameters from its parameter + block (see tramp_table_map()). The trampoline saves the data address in + t0 and jumps to the target code. As ffi_closure_asm() already expects the + data address to be in t0, we don't need a "ffi_closure_asm_alt". */ + +#if defined(FFI_EXEC_STATIC_TRAMP) + .align 16 + .globl trampoline_code_table + .hidden trampoline_code_table + .type trampoline_code_table, @function + +trampoline_code_table: + + .rept 65536 / 16 + pcaddu12i $t1, 16 # 65536 >> 12 + ld.d $t0, $t1, 0 + ld.d $t1, $t1, 8 + jirl $zero, $t1, 0 + .endr + .size trampoline_code_table, .-trampoline_code_table + + .align 2 +#endif + +/* ffi_go_closure_asm. Expects address of the passed-in ffi_go_closure in t2. + void ffi_closure_inner (ffi_cif *cif, + void (*fun)(ffi_cif *, void *, void **, void *), + void *user_data, + size_t *stackargs, struct call_context *regargs) */ + + .globl ffi_go_closure_asm + .hidden ffi_go_closure_asm + .type ffi_go_closure_asm, @function + +ffi_go_closure_asm: + .cfi_startproc + addi.d $sp, $sp, -FRAME_LEN + .cfi_def_cfa_offset FRAME_LEN + + /* Make a frame. */ + SARG $fp, $sp, FRAME_LEN - 2*PTRS + .cfi_offset 22, -2*PTRS + SARG $ra, $sp, FRAME_LEN - 1*PTRS + .cfi_offset 1, -1*PTRS + addi.d $fp, $sp, FRAME_LEN + + /* Save arguments. */ +#if FLTS + FSARG $fa0, $sp, 0*FLTS + FSARG $fa1, $sp, 1*FLTS + FSARG $fa2, $sp, 2*FLTS + FSARG $fa3, $sp, 3*FLTS + FSARG $fa4, $sp, 4*FLTS + FSARG $fa5, $sp, 5*FLTS + FSARG $fa6, $sp, 6*FLTS + FSARG $fa7, $sp, 7*FLTS +#endif + + SARG $a0, $sp, 8*FLTS+0*PTRS + SARG $a1, $sp, 8*FLTS+1*PTRS + SARG $a2, $sp, 8*FLTS+2*PTRS + SARG $a3, $sp, 8*FLTS+3*PTRS + SARG $a4, $sp, 8*FLTS+4*PTRS + SARG $a5, $sp, 8*FLTS+5*PTRS + SARG $a6, $sp, 8*FLTS+6*PTRS + SARG $a7, $sp, 8*FLTS+7*PTRS + + /* Enter C */ + LARG $a0, $t2, 1*PTRS + LARG $a1, $t2, 2*PTRS + move $a2, $t2 + addi.d $a3, $sp, FRAME_LEN + move $a4, $sp + + bl ffi_closure_inner + + /* Return values. */ +#if FLTS + FLARG $fa0, $sp, 0*FLTS + FLARG $fa1, $sp, 1*FLTS +#endif + + LARG $a0, $sp, 8*FLTS+0*PTRS + LARG $a1, $sp, 8*FLTS+1*PTRS + + /* Restore and return. */ + LARG $ra, $sp, FRAME_LEN-1*PTRS + .cfi_restore 1 + LARG $fp, $sp, FRAME_LEN-2*PTRS + .cfi_restore 22 + addi.d $sp, $sp, FRAME_LEN + .cfi_def_cfa_offset 0 + jr $ra + .cfi_endproc + .size ffi_go_closure_asm, .-ffi_go_closure_asm + +#if defined __ELF__ && defined __linux__ + .section .note.GNU-stack,"",%progbits +#endif |